Group Stability and Reading Profiles of Students With Dyslexia: A Double-Deficit Perspective

Abstract

This study examined the Double-Deficit Hypothesis (DDH) by classifying students with dyslexia into four distinct groups, comparing group differences on text-level reading tasks, and examining group stability across one school year (fall to spring). Elementary students (N = 109) were administered measures of reading fluency, reading comprehension, and phonological processing across the school year. DDH group membership was determined by the presence of phonological awareness deficits (PD), naming speed deficits (NSD), double-deficits (DD) in both skills, or no deficits for typically developing (TD) readers. The McNemar test was used to determine the stability of DDH group membership. Analysis of covariance was used to compare DDH groups on text-level reading tasks at each time point after controlling for gender. Overall, reading profiles across the fall DDH groups were congruent with DDH theory, but instability was found in the reading patterns and group membership across time. Nearly half (47.71%) of participants changed DDH groups across the school year, and reading skill differences between the single-deficit groups dissipated in the spring. Results provide partial support for the DDH subgroups. More research is needed to understand the utility of the DDH subtypes for future assessment and intervention practices.

Keywords

double deficit hypothesis assessment reading disabilities identification reading fluency

Wolf and Bowers (1999) developed the Double-Deficit Hypothesis (DDH) to better reflect the multicomponent nature of reading and to understand the deficits related to reading difficulties, specifically dyslexia. According to the Diagnostic and Statistical Manual of Mental Disorders (5th ed.; DSM-5; American Psychiatric Association [APA], 2013), dyslexia is a specific learning disorder (SLD) in reading that is characterized by difficulties in decoding, spelling, accurate word recognition, or fluent word recognition. Prior to the proposal of the DDH, much of the literature focused on deficits in phonological awareness as the primary core deficit associated with dyslexia (Katzir et al., 2008; Kirby et al., 2003). However, dyslexia is a heterogeneous disorder, and the DDH provided a broader lens through which its etiology could be examined. Using the DDH as a framework, the aim of this article was to examine the influence of phonological awareness and rapid automatized naming (RAN) on text-level reading skills in students with dyslexia. Notably, more research was needed in examining the text fluency skills of students with deficits in phonological awareness and RAN over time. Therefore, the literature on these two core deficits areas will be addressed, and then the DDH theory and its hypothesized reading groups reviewed. Finally, the temporal stability of the DDH groups and their respective reading skill profiles (e.g., reading fluency and comprehension) will be examined.

Phonological Awareness and RAN Deficits

The DDH theory posited that deficits in phonological awareness and/or RAN are primarily responsible for the reading difficulties experienced by individuals with dyslexia. Phonological awareness (PA) is an explicit phonological processing skill that requires the student to recognize and manipulate sounds that comprise spoken words (Melby-Lervåg et al., 2012). Students with dyslexia have difficulties in the attainment of these phonological representations. Meta-analytic findings indicated that students with dyslexia tended to perform poorly (two standard deviations lower) on phonological awareness tasks compared with students without dyslexia, and that early phonological awareness skills measured prior to literacy instruction were important predictors of later word reading abilities (Melby-Lervåg et al., 2012; National Institute for Literacy, 2008). During word reading development in kindergarten and first grade, phonological awareness is most needed in the alphabetic phase, when students read both predictable and unpredictable words phonetically (Chall, 1996; Ehri & McCormick, 1998). During second and third grade, decoding common letter patterns becomes automatic, and students develop the ability to read connected text fluently (i.e., accurately, automatically, and with good expression; Kuhn et al., 2010). Yet, even with intact phonological awareness, many students have experienced difficulty in reading text.

RAN, also called naming speed, is a complex skill that requires the integration of many aspects of cognition (Kirby et al., 2010; Norton & Wolf, 2012; Wolf & Bowers, 1999; Wolf et al., 2002). Norton and Wolf (2012) argued that RAN constitutes “a microcosm or mini-circuit of the later developing reading circuitry” (p. 430). Specifically, RAN tasks and reading may require the integration of visual features and pattern information with stored orthographic and phonological representations, leading to later integration and activation of conceptual information, semantic information, and motoric process (Norton & Wolf, 2012; Wolf & Bowers, 1999; Wolf & Denckla, 2005; Wolf et al., 2000). Slowness may be due to difficulty in retrieving phonological stimuli automatically or in sustaining the processes needed for retrieval (Norton & Wolf, 2012; Wolf et al., 2000). Meta-analytic findings of typical and poor readers revealed strong coefficients between RAN and reading fluency across the elementary grades, and moderate correlations between RAN and reading comprehension (Araújo et al., 2015; Swanson et al., 2003). Prior research has postulated that RAN deficits have a larger impact on text-level reading due to the serial, sequential processing demands associated with reading connected text (Kirby et al., 2003, 2010; Norton & Wolf, 2012; Swanson et al., 2003; Wolf et al., 2002). Slow RAN has been thought to restrict the working memory processing needed to recall the meaning between words for comprehension and reading connected text (Norton & Wolf, 2012; Wolf & Bowers, 1999).

Importantly, an integral part of the overall DDH theory is the view of phonological awareness and RAN as related but independent skills with different effects on various reading skills. Prior research demonstrated that RAN accounts for unique variance across text-level reading skills (e.g., reading rate, reading accuracy, reading fluency and reading comprehension) after controlling for phonological awareness which accounts for variance in basic reading skills (decoding and word reading accuracy) (Kirby et al., 2003; Wolf et al., 2002). Particularly, as students have attained automaticity in both phonological awareness and RAN, the focus has shifted away from learning how to read to using their fluent skills to access curricular content (Chall, 1996).

The Double-Deficit Hypothesis (DDH)

Dysfluent reading can be attributed to these specific deficits outlined in the DDH. Thus, readers have been placed into four distinct groups: one typically developing (TD) group, two single-deficit groups, and one DD group (Wolf & Bowers, 1999). Students with intact phonological awareness abilities but who exhibit naming speed deficits are placed in the Naming Speed Deficit (NSD) group (sometimes also called the Rate group). Students with intact naming speed skills but who exhibit deficits in phonological awareness are placed in the Phonological Awareness Deficit (PD) group. The DD group is composed of individuals who exhibit deficits in both areas, whereas those without deficits in either phonological awareness or rapid naming are described as TD readers (Kirby et al., 2003; Lovett et al., 2000; Spector, 2005; Steacy et al., 2014; Wolf & Bowers, 1999; Wolf et al., 2002).

The criteria used for the DDH groups has been relatively consistent throughout the prior literature. Most studies defined a deficit area in phonological awareness or RAN if standard scores were one standard deviation or more below the mean (Katzir et al., 2008; Lovett et al., 2000; Schatschneider et al., 2002; Spector, 2005; Wolf et al., 2002); however, one study used a cut point at the 25th percentile (Steacy et al., 2014). Prior to DDH classification, some studies prescreened students for an SLD in reading based on either a low achievement (e.g., Katzir et al., 2008; Lovett et al., 2000; Wolf et al., 2002) or an ability achievement discrepancy criterion whereas others applied DDH to normative samples (e.g., Kirby et al., 2003; Schatschneider et al., 2002) or to students who were considered at risk for dyslexia based on low performance on measures of decoding, word recognition, and oral reading accuracy (e.g., Spector, 2005; Steacy et al., 2014).

Using the DDH criteria, researchers were able to classify the majority of participants into one of the DDH groups (73%–84%), and students with severe reading disabilities most often meet criteria for the DD group (23%–60%) (Katzir et al., 2008; Lovett et al., 2000; Spector, 2005; Wolf et al., 2002). Steacy et al. (2014) reported the lowest percentage of DD students, with only 13% of students classified as DD at the beginning of kindergarten, which makes sense given that participants were considered at risk for a reading disability. Studies of normative readers yielded higher percentages for TD group and lower for the single-deficit groups and DD groups (Kirby et al., 2003; Schatschneider et al., 2002). Different trends in DDH membership emerged across different grades, and across studies with samples consisting of at-risk readers versus students with severe reading disabilities. Among students identified as at risk for dyslexia at the beginning of kindergarten, group membership was 14% for PD group, 14% for NSD group (NSD), 13% for DD group, and 59% for TD group (Steacy et al., 2014); however, at the beginning of first-grade, group membership was 33% for PD, 17% for NSD, 23% for DD, and 27% for TD (Spector, 2005). Among studies of students identified as having severe difficulty with reading in second and third grade, researchers found the highest group membership among DD groups was 60% (Wolf et al., 2002) and 46% (Katzir et al., 2008).

Reading Skill Profiles of DDH Groups

The performance of the DDH groups has been examined across a wide array of reading skills including decoding (e.g., pronouncing pseudo words), word reading (e.g., identifying regular and irregular sight words), reading fluency (e.g., passage reading rate and accuracy), and comprehension (Katzir et al., 2008; Kirby et al., 2003; Lovett et al., 2000; Schatschneider et al., 2002; Wolf & Bowers, 1999; Wolf et al., 2002). Differences across the DDH groups were most apparent as tasks transitioned from lower level skills (letter and word) to reading connected text fluently and with good comprehension. For differences in terms of reading comprehension among single-deficit groups, the NSD group typically outperformed the PD group (Lovett et al., 2000; Schatschneider et al., 2002), although group differences were not always apparent (Katzir et al., 2008). Studies have consistently found DD readers having severely impaired comprehension compared with the other groups (Katzir et al., 2008; Kirby et al., 2003, 2010; Lovett et al., 2000; Schatschneider et al., 2002; Spector, 2005; Wolf & Bowers, 1999; Wolf et al., 2002). This pattern is not surprising as the combined impact of phonological awareness and RAN deficits influence multiple aspects of the reading process (Araújo et al., 2015; Kirby et al., 2003, 2010; Norton & Wolf, 2012; Schatschneider et al., 2002; Swanson et al., 2003). These patterns suggest phonological awareness and RAN are both important predictors of reading skill, and students with double deficits in both areas are likely to experience the most severe difficulties in reading text.

Unlike reading comprehension, to our knowledge only one cross-sectional study has examined the performance of DDH groups on individual measures of reading fluency. Katzir et al. (2008) parsed reading fluency into rate and accuracy, noting that combining these skills into a more global metric risked obfuscating important aspects of text-level reading. Across both rate and accuracy, the phonological deficit (PD) group outperformed the other two groups; however, the DD and NSD groups were not found to differ. Rate and accuracy scores fell in the below average range for all three DDH groups, which is not surprising given the sample consisted of students with dyslexia. To explain these differences among groups, Katzir et al. (2008) theorized the DDH groups could have unique paths that lead to dysfluent reading. For the NSD group, it was argued that slow rate across levels of reading (letter, sentence, and text) interfered with reading accuracy. Conversely, it was argued that deficits in accuracy slowed reading rate for the PD group. The DD group had the worst prognosis for reading with impaired reading rate and accuracy at each reading level and lower verbal skills (Katzir et al., 2008). However, 39% of participants in the sample, mainly from the DD group, were missing reading fluency data because the task was too challenging for them to complete, which complicated the interpretation of these group differences. With few longitudinal studies examining DDH groups on performances of reading fluency, more research is needed to further support Katzir’s claim of unique paths leading to dysfluent reading.

Temporal Stability of DDH Groups

Only a handful of researchers have explored the temporal stability of the DDH groups (e.g., Spector, 2005; Steacy et al., 2014). In prior longitudinal studies, subtyping was typically done either at pretest or posttest rather than at both time points (e.g., Kirby et al., 2003; Lovett et al., 2000; Schatschneider et al., 2002). Spector (2005) classified 196 first-grade students who were at risk for dyslexia into one of the DDH groups and examined the stability of group membership from the fall to the spring of one school year. A total of 80% of first grade students were identified with a single deficit in phonological awareness or RAN in the fall. Yet at both time points, only about 50% of the students maintained their group membership. Only 65% of the PD group, 30% of the NSD group, and 40% of the DD group were consistently in the same group from fall to spring. Kirby et al. (2010) noted the instability of the groups in the Spector (2005) study were perhaps due to the developmental period the groups were formed (e.g., first grade) or the presence of interventions. Indeed, 106 students in Spector (2005) study were drawn from a reading recovery intervention study. In addition, Steacy et al. (2014) hypothesized these transitions into different groups could be due to unreliable measurement or to some students failing to benefit from instruction compared with their peers.

To examine stability further, Steacy et al. (2014) followed 158 students at risk for phonological awareness and RAN deficits from the fall of kindergarten through the spring of their second-grade year. Across groups, the largest transition was found to occur from the spring of kindergarten to the fall of first grade. The majority of these transitions were students moving from a single-deficit group (NSD or PD) to the DD group. From kindergarten to first grade, students in the PD group had a probability of 29% and students in the NSD had a probability of 34% of moving into the DD group. Only 8% of the TD members transitioned into either PD or NSD from kindergarten to first grade, while 19% of the DD group transitioned to either the PD or NSD groups. Interestingly, groups were found to be most stable during the second-grade year (fall to spring). In the fall of second grade, the PD group had a 97% likelihood, the NSD group had an 85% likelihood, and the DD group had an 89% likelihood of maintaining their group membership into the spring. The results from this study suggest that deficits may stabilize over time, but more research is needed to further support that claim.

These two longitudinal studies also examined differences in reading performance across the groups classified by the DDH. When looking at the reading profiles of DDH groups across the first-grade year, Spector (2005) found that the DD readers performed significantly lower than the other groups on text-level reading accuracy in the spring. However, Spector (2005) did not find significant differences between DD readers and the other three groups in the fall. The authors noted this lack of difference could be due to so many students scoring at floor level on the measures of reading, or the participation in a short-term reading intervention which could have impacted performance (Spector, 2005). In the spring of first grade, the NSD group performed better than the DD group on word attack and passage comprehension measures. However, in the fall of second grade, both the phonological awareness and NSD groups performed better than the DD group on the same measures. These trends are consistent with prior research that points to the differential influence of phonological awareness and naming speed on various reading tasks, and the declining performance of the DD group compared with the other groups in later grades (Katzir et al., 2008; Kirby et al., 2003, 2010; Schatschneider et al., 2002; Spector, 2005; Steacy et al., 2014; Wolf & Bowers, 1999; Wolf et al., 2002). Given the scant literature, the reading skill profiles and temporal stability of DDH groups need to be further examined longitudinally and with a clinical sample inclusive of a broader range of grades.

Purpose of the Study

The purpose of this study was to examine the temporal stability of DDH group membership, and the reading skill profiles displayed across these groups over the course of a school year in elementary students with dyslexia. As previously noted by Steacy et al. (2014), although a myriad of studies has examined reading profiles, no study to date has examined the temporal stability of the DDH groups in a clinical sample. The handful of prior studies of this topic focused on early elementary students, ranging from kindergarten to second grade, and therefore focused primarily on word-level measures of reading skills (Spector, 2005; Steacy et al., 2014). Little research has examined how deficits in phonological awareness and RAN impact higher level reading skills, such as reading fluency and comprehension, that are necessary for academic success in late elementary school and middle school years (Katzir et al., 2008; Lovett et al., 2000; Schatschneider et al., 2002; Steacy et al., 2014; Wolf et al., 2002). This study expanded the literature by examining how patterns across RAN and PDs influence text-level reading skills (e.g., fluency and comprehension) across multiple grade levels. Consistent with Katzir et al. (2008), reading fluency was parsed into rate and accuracy to allow for a more nuanced examination of text-level reading skills.

Using the DDH as a framework for this study, a clinical sample (i.e., all participants were diagnosed with dyslexia) of first- through fifth-grade students were categorized into the PD, rapid naming deficit (NSD), DD, or typical developing (TD) reader group. The following research questions guided this study: (1) Can the DDH groups be replicated in a clinical sample of students with dyslexia? (2) How stable is DDH group membership from the beginning to the end of the school year (August to May)? (3) Do the reading profiles of DDH groups differ in terms of rate, accuracy, and reading comprehension skills? (4) How stable are the reading profiles of the DDH groups from the beginning to the end of the school year (August to May)? Based on the prior research, it was hypothesized that the majority of students would be categorized into one of the DDH groups, and that the DD group would have the most members. Based on the limited prior literature, DDH group membership was expected to have little instability with some transitions from single-deficit groups into either the DD or typical developing group from fall to spring. Finally, differences in the reading profiles of the DDH groups were expected. Students in the DD group were expected to perform poorly across all text-level tasks compared with the other groups and across both time points. Different patterns of performance were expected across the single-deficit groups based on the nature of their respective deficits. Specifically, the PD group was expected to outperform the rapid naming deficit group on passage rate and accuracy tasks while the rapid naming deficit group was expected to outperform the PD group on reading comprehension.

Method

Participants

Students were recruited from a local private elementary school that specializes in serving students with dyslexia. Participants were 109 students in first grade (n = 15), second grade (n = 24), third grade (n = 23), fourth grade (n = 23), and fifth grade (n = 24). Students were approximately 88.2% European American, 6.4% African American, 3.6% Other, 0.9% Asian, and 0.9% Hispanic; 60% were boys. On average, participants were 108.00 months old (SD = 17.40, range 75.00–144.00) at date of fall testing. Based on available information from school records, the average full-scale IQ for the sample fell in the average range (M = 102.39; SD = 12.66). School admission required a diagnosis of an SLD with impairment in basic reading skills (i.e., dyslexia) using DSM-5 criteria (APA, 2013). All participants demonstrated impairments in the subskills of word reading accuracy and/or reading rate or fluency, and many experienced concurrent impairments in reading comprehension. As would be expected in a clinical sample, many students had an educationally relevant comorbid diagnosis, that is, SLD in written expression 35%, SLD in mathematics 15%, attention-deficit/hyperactivity disorder (ADHD) 45%, and speech and/or language impairment 19%.

Participants were selected from a longitudinal study examining the development of reading fluency and related skills among students with dyslexia across three academic years (2015–2016; 2016–2017; 2017–2018). Many participants had multiple years of data available, but only one academic year was selected for use in this study. If a given student had more than 1 year of data available, the year with complete data for the variables of interest in this study was selected. If complete data were available for multiple years (which was largely the case as instances of missing data were quite rare), then the grade with the lowest number of participants was selected to create a more even distribution across grade levels. Using effect sizes reported in the previous literature (i.e., Steacy et al., 2014), a power analysis determined that a minimum sample size of 48 students was needed for the proposed analysis of covariance (ANCOVA) (G*Power 3; Faul et al., 2007). Therefore, the size of the available sample (N = 110) should allow for adequate statistical power.

Measures

Selected subtests from the Comprehensive Test of Phonological Processing, Second Edition (CTOPP-2; Wagner et al., 2013) were administered to assess phonological awareness and RAN skills, and to the resulting standard scores (M = 100; SD = 15) were used to place students into DDH groups. The Elision subtest was used to measure phonological awareness, and the Rapid Letter Naming and Rapid Number Naming subtests were used to measures RAN. The Elision subtest consists of 34 items that measure the student’s ability to remove some sounds (phonological segments) from words to form other words. For example, the examinee is instructed to say “sunshine.” After repeating the word “sunshine,” the examinee is instructed to say “sunshine” without saying “sun.” The correct response would be “shine.” The items increase in difficulty by asking the examinee to remove specific sounds. For the Elision subtest the test–retest reliability coefficients were .93 for ages 4 to 6-year-old students and .77 for 7 to 11-year-old students, the internal consistency reliability coefficient was .91, and the average concurrent validity estimate for 4- to 18-year-old students was .70 (Wagner et al., 2013).

The Rapid Symbolic Naming composite consists of the Rapid Digit Naming and Rapid Letter Naming subtests. Both subtests measure the student’s ability to retrieve phonological and visual information from long-term memory. The Rapid Digit Naming subtest requires students to name randomly arranged numbers presented in rows on a page. The score is based on the number of seconds it takes for the student to name all the numbers on the page. The Rapid Letter Naming uses the same format and procedures but requires the student to name letters rather than numbers. According to the manual, test–retest reliability coefficients were .81 for 4- to 6-year-old students and .89 for 7- to 11-year-old students, the internal consistency reliability coefficient was .92, and the average concurrent validity estimate for 4- to 20-year-old student was .45 (Wagner et al., 2013).

The Gray Oral Reading Tests, Fifth Edition (GORT-5; Wiederholt & Bryant, 2012) assesses various aspects of oral reading skills at the connected text level, namely rate, accuracy, reading fluency, and reading comprehension. This test consists of 16 passages of increasing in difficulty that are read aloud until the ceiling criterion is reached. Reading time and errors are recorded, and five open-ended comprehension follow each passage. Students were administered form A in the fall and form B in the spring. The Accuracy score is based on the number of word reading errors for each passage (i.e., omissions, substitutions, insertions, repetitions, self-corrections, and reversals). The Rate score is based on the time in seconds taken to read each passage. Rate and Accuracy combine to produce the Fluency subtest. The Comprehension score is based on the number of questions about the stories that the student answers correctly. These passage scores are then summed to generate a scaled score (M = 10, SD = 3) for each subtest. According to the manual, average coefficient alphas ranged from .92 to .94, test–retest reliability ranged from .82 to .90, and alternate-form reliability ranged from .77 to .88 for these subtests; average correlations between the GORT-5 subtests and other reading measures ranged from .68 to .74 (Wiederholt & Bryant, 2012).

Procedure

Written parental consent and student assent were a requirement for participation in this study on an annual basis. Three waves of data were collected during the 2015–2016, 2016–2017, and 2017–2018 school year. All measures were administered by examiners with previous training and experience in psychoeducational assessment. Prior to data collection, each examiner practiced coding the assessments with audio recordings of students’ reading until a minimum of 95% agreement was reached. Administration of the GORT-5 and selected subtests from the CTOPP-2 were conducted over an approximately 2-week period at the beginning (August) and end (May) of each academic year. The measures used in this study represent a subset of the tests administered as part of a larger study. The administration order of the assessments was counter-balanced within each time point to control for order effects. Measures were individually administered in an empty classroom or office during regular school hours. All testing sessions were audio recorded for later scoring. At the end of each session, students received a prize, such as a sticker or small toy slinky, in appreciation for completing the assessments. After the data were collected, an independent reviewer selected 25% of the data for blind review using an online randomizer to ensure interrater agreement. Discrepancies were resolved via discussion and rarely occurred (<1%) for the data collected in Waves 1, 2, and 3.

DDH Group Membership

To determine DDH group membership at each time point, scores from the CTOPP-2 Elision subtest and the Rapid Symbolic Naming composite were used as criterion variables for phonological awareness and RAN. To be consistent with the prior literature, the below criterion for group membership was used in this study. To qualify for the PD group, participants had to score at or below the 16th percentile (i.e., one standard deviation below the mean) on the Ellison subtest (scaled scores ≤ 7) but above the 16th percentile on the Rapid Symbolic Naming composite (standard scores ≥ 85). Similarly, to qualify for the NSD group (ND), participants had to score at or below the 16th percentile on the Rapid Symbolic Naming composite but above the 16th percentile on the Elision subtest. Participants with scores at or below the 16th percentile in both phonological awareness and rapid naming skills were placed in the DD group, whereas those with scores above the 16th percentile in both areas were placed in the TD reader group.

Results

Data Screening and Analytic Plan

The DDH criteria were applied to the data and four distinct groups emerged in the fall and spring. In addition, the stability of group memberships from the beginning to the end of the school year was examined using McNemar’s chi-square test, a nonparametric approach used to determine if there is significant change among matched paired data (Vuolo et al., 2015). Finally, to examine group differences on various aspects of text-level reading (i.e., rate, accuracy, and comprehension), a series of one-way ANCOVAs was conducted separately by time point (i.e., fall and spring) with age and gender serving as covariates.

Prior to analyses, the data were screened for accuracy of input, implausible or out of range values, outliers (z-scores < 3.29), and normality (i.e., skewness and kurtosis < |2.0|) (Tabachnick & Fidell, 2013). No missing data or outliers were identified, and skewness and kurtosis statistics fell within the appropriate range (see Table 1). As would be expected for a clinical sample of students identified with dyslexia, measures revealed significant reading difficulties in the fall while gains were made in the spring.

Table 1.

Descriptive Statistics for the Fall and Spring (N = 109).

Measure	Fall				Spring
Measure	M	SD	Skewness	Kurtosis	M	SD	Skewness	Kurtosis
CTOPP-2 Elision	8.10	2.39	−0.03	−0.33	8.88	2.66	0.16	0.85
CTOPP-2 Rapid Symbolic Naming	87.8	12.25	0.36	0.65	91.6	11.09	0.06	0.31
GORT-5 Accuracy	5.66	1.74	0.31	−0.67	6.44	1.93	−0.04	−0.93
GORT-5 Rate	6.38	1.91	−0.03	−0.68	7.37	1.99	−0.16	−0.40
GORT-5 Comprehension	6.25	1.68	−0.04	−0.34	7.33	1.76	0.06	0.13

Note. CTOPP-2 = Comprehensive Test of Phonological Processing, Second Edition; GORT-5 = Gray Oral Reading Tests, Second Edition.

Other assumptions required by ANCOVA were examined. The independence of the covariate and independent variable was tested using ANCOVA with gender and age as the outcome and the DDH groups as the predictor. Age was not significantly different across the four DDH groups in the fall or spring. Gender was significantly different in the fall among the DDH groups, F(1, 105) = 5.863, p= .017; however, gender differences were not found among the spring DDH groups. Since this assumption was violated for the fall time point, one option was to remove gender as a covariate. However, removing gender as a covariate in the fall did not change the results, so it was kept for the sake of consistency. Heterogeneity of regression slopes becomes a concern if there is an interaction between the independent variables and the covariates (Tabachnick & Fidell, 2013). However, neither of the covariates used in this study (gender and age in months) were found to interact with DDH group in either the fall or the spring (all p >.05). Levine’s test was found to be significant for one ANCOVA (p< .05), when accuracy served as the dependent variable in the fall. It should be noted that Levene’s test of homogeneity is overly conservative, meaning that small differences in group variances can produce significance when in fact the assumption of homogeneity has not been violated. Therefore, Hartley Fmax was examined, and results suggested that homogeneity was not violated for fall accuracy

Group Membership

To address the first research question (i.e., can the DDH groups be replicated in a clinical sample of students with dyslexia), the DDH group membership criterion were applied to the data (see Table 2). All 109 of the students were classified using the DDH criterion, resulting in the following group memberships: TD group (n = 38), PD group (n = 17), NSD group (n = 27), and DD group (n = 27).

Table 2.

Temporal Stability of Double-Deficit Hypothesis Groups (N = 109).

Fall	Spring
Fall	TD (n = 56)	PD (n = 18)	NSD (n = 15)	DD (n = 20)
TD (n = 38)	31	4	1	2
PD (n = 17)	5	7	2	3
NSD (n = 27)	12	1	9	5
DD (n = 27)	8	6	3	10

Note. TD = typically developing; PD = phonological awareness deficit; NSD = naming speed deficit; DD = double-deficit.

Temporal Stability

Once groups were established, the stability of the DDH group membership across the school year was examined. In this sample, 47.71% of students transitioned to different DDH group at the end of the school year. To better understand these changes in group membership, McNemar’s test for the homogeneity of proportions was utilized. Each of the DDH groups was tested separately, as McNemar’s test examines change in binary repeated measurement. Group membership in the fall was compared with their equivalent group in the spring (i.e., fall PD vs. spring PD). A significant difference in the proportion of group members from the fall to the spring was not found for the PD group (p = 1.00) and the DD group (p = .248). However, there was significant change in group membership for the NSD group (p = .023) and the TD group (p = .003) from the fall to the spring. In other words, there was a similar proportion of students in the PD group and DD group at the beginning and end of the school year. In contrast, the NSD group had a lower proportion of members and the TD group had a higher proportion of members at the end of the school year.

A closer examination of these transitions revealed that 82% of those in the TD group in the fall maintained their membership in the spring, 3% declined in RAN skills and moved to the NSD group, 10% declined in phonological awareness skills and moved to the PD group, and 5% declined in both skills and moved to the DD group. Not only did the TD group demonstrate the greatest stability across the year, but it also gained the most new members with the majority of the spring transitions (25 out of 52) from the deficit groups occurring into the TD group. In contrast, the NSD group had the lowest stability with only 33% of members retained in the spring, 44% improved in RAN skills and moved to the TD group, 18% declined in phonological awareness skills and moved to the DD group, and 5% transferred to the other single-deficit group (NSD to PD) in the spring. For the fall PD group, 41% were retained in the spring, while 29% improved in phonological awareness skills and moved to the TD group, 13% switched single-deficit groups (PD to NSD), and 17% showed a decline in RAN skills and moved to the DD group. For the fall DD group, 37% of members were retained in the spring, 11% improved in RAN skills and moved to the PD group, 22% improved in phonological awareness skills and moved to the NSD group, and 30% improved in both skills and transitioned to the TD group in the spring. Importantly, 34 of the 52 students who changed groups across the year moved into a less impaired group (i.e., from DD to single deficit, from single to TD).

Reading Profiles

A series of one-way ANCOVAs were conducted to examine whether DDH groups (TD; PD; NSD; & DD) differed in terms of their reading performance on the GORT-5 Accuracy, Rate, and Comprehension subtests (see Table 3). Given that nearly half of all students transitioned to a new DDH groups across the school year, it was necessary to analyze the fall and the spring data separately. To reduce the probability of Type I error associated with multiple comparisons, a Bonferroni correction was applied resulting in adjusted alpha level of .008 (.05/6).

Table 3.

Descriptive Statistics and Planned Comparisons: Analysis of Covariance.

Measures		Fall					Spring
Measures		PD	NSD	DD	TD	Planned comparisons	PD	NSD	DD	TD	Planned comparisons
Accuracy	n	17	27	27	38	DD < PD & TD NSD < TD	18	15	20	56	DD < TD**
	M	6.35	5.22	4.44	6.53		6.06	6.13	4.95	7.18
	SD	1.58	1.28	1.22	1.83		1.98	1.77	1.43	1.78
Rate	n	17	27	27	38	NSD & DD < PD & TD**	18	15	20	56	DD < TD &PD**
	M	8.06	5.70	5.22	6.92		7.67	6.73	5.70	8.04
	SD	1.25	1.73	1.69	1.72		1.53	1.91	1.78	1.86
Reading Comprehension	n	17	27	27	38	DD < PD & TD**	18	15	20	56	DD < TD**
	M	7.29	6.15	5.19	6.61		6.94	6.73	6.40	7.95
	SD	1.49	1.61	1.27	1.70		1.63	1.03	1.57	1.82

Note. Gender and age in months served as covariates. PD = phonological awareness deficit; NSD = naming speed deficit; DD = double-deficit; TD = typically developing.

p < .05. **p < .01.

At the fall time point, DDH group membership had a significant effect on reading rate even after controlling for gender and age, F(3, 105) = 12.51, p < .001, partial η² =.267. Pairwise comparisons indicated that the PD and TD groups performed significantly better than both the NSD group and the DD group on the measure of rate (p < .008; see Table 3). The DDH groups were also found to differ in terms of their reading accuracy after controlling for demographic variables, F(3, 105) = 12.31, p < .001, partial η² =.264. Pairwise comparisons indicated the DD group performed significantly worse than both the PD and TD group on reading accuracy, whereas the TD group also performed significantly better than the NSD group (p < .008; see Table 3). Finally, reading comprehension was also found to vary across the DDH groups after controlling for demographic variables, F(3, 105) = 7.44, p < .001, partial η² =.178. Pairwise comparisons revealed that the PD and TD group performed significantly better on reading comprehension than the DD group (p < .008; see Table 3). Importantly, the impact of DDH group membership on text reading skills were relatively robust, with a large effect sizes for rate and accuracy and a medium effect size found for reading comprehension using common guidelines for interpretation (<.09 small, .09–.24 medium, >.25 large; Cohen, 1988).

The final research question addressed whether the reading profiles of the DDH groups were stable across the school year (August to May). Due to the high frequency of transitions in DDH membership across the year, time could not be examined as an independent variable in the reading profiles analyses. Therefore, separate ANCOVAs were also conducted for the spring. Congruent with the fall results, spring DDH group membership had a significant effect on reading rate, F(3, 105) = 8.496, p < .001, partial η² =.198, reading accuracy, F(3, 105) = 7.793, p < .001, partial η² =.185, and reading comprehension performance, F(3, 105) = 5.088, p =.003, partial η² =.129. However, the results were somewhat less pronounced, with effect sizes falling in the medium range for all three aspect of text reading. In the spring, pairwise comparisons indicated that the DD group performed significantly worse than the both PD and the TD group in terms of reading rate (p < .008; see Table 3). However, on both reading accuracy and comprehension, results indicated that only the DD group performed significantly worse than the TD group (p < .008; see Table 3). Across both time points, the only instance where a covariate was found to be significant was found for gender on fall reading accuracy, F(1, 105) = 23.922, p < .001, partial η² =.188, with boys (M = 6.22) outperforming girls (M = 4.84).

Discussion

To our knowledge, this is the first study to examine the stability of DDH group membership and their associated reading profiles across the school year in a clinical sample of students diagnosed with dyslexia. All students were classified into one of the DDH groups and reading profiles congruent with the literature were apparent in the fall. However, the changes in both group membership and reading patterns displayed in the spring were not consistent with the study’s initial hypothesis. Consequently, the results of this study provided partial support for the DDH theory for classifying students into reading subtype groups.

Consistent with the DDH theory and prior literature, the DDH groups were replicated in across the fall and the spring time points. As the DDH was meant to be applied to children with dyslexia, this result was not surprising. At the beginning of the school year, the PD group compromised 15%, the NSD and DD groups each represented 25%, and the TD group compromised 35% of the sample. These percentages differed somewhat from other studies that examined DDH membership at single time point in clinical or at-risk samples, where the DD group was found to represent the largest proportion (46%-60%) of their sample (Katzir et al., 2008; Lovett et al., 2000; Wolf et al., 2002). This variation in the prevalence of the DD group could be due the prescreener used in some studies prior to data collection to select only those with severe reading impairments (e.g., Katzir et al., 2008; Lovett et al., 2000; Wolf et al., 2002). Lovett et al. (2000) applied a prescreener criterion of scores falling below the 20th percentile on five standardized reading achievement measures to a clinical sample of students with severe reading disabilities. Both Wolf et al. (2002) and Katzir et al. (2008) examined the differences among DDH groups in low achievement and ability achievement discrepant groups of students in second and third grade with severe reading impairments. In contrast, all students at the participating school were eligible for participation in this study and, based on the percentage of TD readers, it is likely that our sample included individuals whose reading deficits were partially remediated. However, it is important to note that our results from the fall time point in terms of group percentages were generally consistent with the two studies that examined temporal stability (i.e., Spector, 2005; Steacy et al., 2014).

In contrast, the resulting temporal stability of the DDH groups was not congruent with our hypotheses. Nearly half (47.71%) of students in this study transitioned into a different group at the end of the school year. The majority of these transitions were into the TD group, suggesting that students were benefiting from the intervention provided. Similar to the current study, Spector (2005) examined the stability of DDH groups across two times points and also found relatively low stability (i.e., 50%) in DDH group membership across the school year. However, stability was examined across intervention participants and wait-listed first-grade students. Due to this instability, Spector investigated whether the early intervention explained these transitions but did not find stability to vary across intervention and wait-listed students.

Steacy et al. (2014) found that groups demonstrated some instability between their first two time points (spring of kindergarten to the fall of first grade) as evidenced by the lower probabilities of maintaining membership (ranging from .08 to .44). However, from the fall to the spring of first grade and the spring of first grade to the fall of second grade, transition probabilities revealed robust stability (ranging from .71 to 1.00) with groups showing increased stability each year. Both Spector (2005) and Steacy et al. (2014) found that the NSD group had the fewest members and lowest probability of maintaining membership (.56 and .68, respectively). Although more than 70% maintained the same classification from kindergarten to second grade, the TD group (48%) had the largest proportion of students who demonstrated stability while only 10% of the PD, 7% of DD, and 5% of the naming speed groups were stable (Steacy et al. 2014). Results from the current study revealed a similar trend in that the naming speed group maintained the lowest percentage (33%) whereas the TD group maintained the greatest (82%) percentage of members across the year. Likely due to improvements in skills across the year, the TD group in this study also gained the most members accounting for the majority of the spring transitions. To create nuanced groups, Steacy et al. (2014) oversampled for children at risk for dyslexia by excluding students with standard scores near the normed population means to eliminate extreme difference in phonological awareness and RAN among the groups. In addition, students in the TD group with high scaled scores on phonological awareness (>9) and RAN (>11) were dropped from the sample. The 25th percentile cutoff for placement in the DDH group used by Steacy et al. (2014) was based on the sample rather than established norms, whereas this study used a 16th percentile cutoff using norm-referenced scores. This precise and nuanced criterion applied to eliminate extreme differences in the Steacy et al. could explain the robust findings of stability by reducing likelihood of transitions and variability.

Across the studies examining the temporal stability of the DDH groups, clear trends emerged regarding which groups were the most and least stable across time (i.e., the TD and NSD groups, respectively). Although variation in the overall stability may be due to methodological differences in establishing groups, a variety of factors may have contributed to the temporal instability of DDH groups found here and by Spector (2005). One explanation is the arbitrary nature of the cutoff scores that were used, which could misclassify students with scores on the Elision subtest and Rapid Symbolic Naming composite just above or below the cutoff criterion. Relatively small changes in scores across the year due to improvement in skill or measurement error could result in a change in group membership status. Although Spector did not find support for early intervention attributing to instability, it is possible use of a rigorous intervention by the school participating in this study could have influenced the progression of higher scores on both phonological awareness and RAN tasks. Phonological awareness has been proven to be readily malleable with intervention (Schwanenflugel & Knapp, 2016) and in Vukovic and Siegel’s (2006) analysis of the literature found that RAN improved with phonological based intervention. Indeed, this interpretation is supported by the pattern of transitions across DDH groups from greater to lesser impairment. Stability in DDH group membership for those receiving high-quality intervention services may not be a reasonable expectation.

To explore group differences on various aspects of text-level reading (i.e., rate, accuracy, and comprehension), the reading profiles associated with each DDH group were examined in the fall and the spring separately. As hypothesized, the fall profiles were consistent with those found by Katzir et al. (2008) in that the PD group outperformed the NSD group on measures of rate, and the single-deficit groups did not differ in passage comprehension. However, unlike previous findings, the PD group did not outperform the naming speed group on measures of accuracy in the fall. This result is surprising given that prior research has found naming speed to be a strong predictor of reading fluency skills, particularly reading rate (Kirby et al., 2003, 2010; Norton & Wolf, 2012; Swanson et al., 2003; Wolf et al., 2002). Consistent with the previous literature, the DD group had the lowest scores on all reading tasks and significantly differed from the phonological deficit and TD groups (e.g., Katzir et al., 2008; Kirby et al., 2003, 2010; Lovett et al., 2000; Schatschneider et al., 2002; Spector, 2005; Wolf & Bowers, 1999; Wolf et al., 2002). In addition to this finding, there were no significant differences found in performance between the DD and the naming speed groups, but this pattern is consistent with another study that examined text-level reading skills (Katzir et al., 2008).

In the spring, the group differences between the single-deficit groups and the DD group largely dissipated. The DD group still performed significantly worse than the PD group on measures of rate, but only performed significantly worse than the TD group on measures of accuracy and passage comprehension. The phonological awareness and NSD groups did not significantly differ from each other on any of the reading performance tasks. These results contradict Spector’s (2005) finding from the spring and Steacy et al. (2014) findings in the spring of first grade and the fall of second grade that the DD performed significantly worse than both the naming speed and the PD groups on passage comprehension tasks. However, it should be noted that those two longitudinal studies primarily examined word-level tasks such as letter identification, word identification, and pseudoword reading, whereas this study focused on text-level reading. Given the poor temporal stability found for group membership, the differences in reading profiles for the DDH groups across the year are not surprising as the spring groups are composed of different members than in the fall.

Theoretical Implications

Results from this study offer partial support for the DDH theory. The DDH criteria were successful in categorizing all students into one of the four groups, and initial group performance trends in the fall were in the expected direction. Therefore, these results align with DDH theory in terms of the impact deficits of phonological awareness and RAN deficits on text-level reading skills when examined at a single time point at the beginning of the year. DDH theory was pivotal is drawing attention to the role of RAN in the reading literature, and these results do not undermine that contribution. However, the poor temporal stability of the DDH groups suggest that the subtyping students based on these deficits may not be pertinent for intervention planning purposes. As previously noted, the expectation of group stability in the face of high-quality intervention may not be tenable. Furthermore, past research has found cause to question the reliability and validity of the DDH groups (Compton et al., 2001; Schatschneider et al., 2002; Vellutino et al., 2004; Vukovic & Siegel, 2006).

Notably, methodological problems have been associated with the formation of the four DDH subgroups based on phonological awareness and RAN scores. Due to the fact that phonological awareness and RAN are continuous measures that typically correlate with one another, and because the relation between phonological awareness and reading performance tasks tend to be curvilinear, Schatschneider et al. (2002) argued that students in the DD group would be expected to have considerably lower scores in phonological awareness compared to the PD group. Therefore, large differences would be attributed primarily to deficits in phonological awareness and related phonological skills and not to the combined effects of both phonological awareness and RAN, a central tenant of the DDH theory (Vellutino et al., 2004). A series of post hoc t tests was used to explore this explanation. Still, the DD group was not found to score lower on phonological awareness and RAN compared with either of the single-deficit groups at both time points (p > .05).

A major finding from the current study is the lack of unique differences between the deficit groups on reading tasks in the spring. Compton et al. (2001) found that differences on reading comprehension among the DD group and the single-deficit groups disappeared when the groups were matched in scores on phonological awareness and RAN. Because our study found similar results to Compton et al. (2001), this finding could explain the lack of differences on reading performance task in the spring. One common statistical issue that arose with Compton et al. (2001) study is the use of a cutoff criterion procedure, which resulted in the creation of mismatch between the DD group and single-deficit group. One way to address this statistical issue would be to use a buffer zone between groups for students who perform below average on tasks of phonological awareness and RAN and close to TD based on standard scores. Another methodological issue raised by Vukovic and Siegel (2006) was the inconsistent existence of an NSD group and associated questions surrounding whether naming speed difficulties demonstrate a predictive and stable relation with reading ability. The current study was able to replicate an NSD group in both the fall and the spring, but the NSD group had the lowest stability across the year in this study and prior literature (Spector, 2005; Steacy et al., 2014). Although results from this study point to the existence of both concurrent and single phonological awareness and RAN deficits, the temporal stability of groups based on deficit patterns in those areas was not supported. Thus, the potential methodological issues create potential barriers in the interpretation of the DDH by raising questions regarding whether differences among groups are attributed to significant deficits or measurement error.

Due to the poor temporal stability of the DDH groups observed in this study and elsewhere (Spector, 2005), the applicability of these subtypes may be limited in practice, especially in terms of informing intervention efforts. Temporal variability could be accounted for a variety of reasons such as metalinguistic components for phonological awareness, speech-motor articulation for RAN, instructional factors, and other individual differences in phonological coding and phonological access (Norton & Wolf, 2012; Vellutino et al., 2004). Indeed, after reviewing the literature and in light of the current study, it is difficult to find commonality among the theoretical, methodological, psychometric standards, and results from studies examining the DDH groups. Though the DDH provides a deeper understanding of the components of reading, the utility of DDH groups in strengthening our assessment and intervention efforts for students who struggle in reading is not clear. Therefore, it is difficult to draw conclusions regarding how to integrate our findings with the current literature, especially with limited longitudinal studies.

Limitations

The current study had some notable methodological strengths including a lack of attrition, no missing data, a unique clinical population, and the detailed attention to data quality issues through the data collection process (i.e., counterbalancing and inner-rater reliability procedures). However, several limitations of this work warrant discussion. The Elision subtest was selected to measure phonological awareness to be consistent with prior literature (i.e., Katzir et al., 2008). Although this measure has strong psychometric properties, the use of a single subtest may have resulted in a less reliable estimate of skill than if a composite score had been used. The homogeneity of the sample may limit the generalizability of these results. The majority of participants were White and from middle-class backgrounds, and this sample may not reflect the diversity seen in many communities. These students do not represent the normal distribution of reading skills, but rather the lower end of the distribution. Because participants attended a private school tailored to serving individuals with dyslexia, they likely received more intensive instruction than would be seen in a typical school setting. Finally, the size of our sample and the use of a short-term longitudinal design (i.e., two time points across a single year) limit the conclusions that may be drawn regarding developmental trends and long-term growth.

Implications for Future Research and Practice

Future studies should incorporate analysis of larger, more diverse clinical samples across multiple years to replicate the instability of the DDH groups and their associated reading profiles. In addition, more research is needed to see how deficits in RAN and phonological awareness in the context of the DDH groups affect higher demanding reading tasks, especially text-level reading fluency. The current study was novel in that it parsed fluency into rate and accuracy and examined these skills into the upper elementary grades; however, one important component of fluency, reading prosody, was not addressed and should be examined in the future. Perhaps the most impactful outcome from the DDH theory is the increased focus on RAN in both research and clinical endeavors. Both phonological awareness and RAN assessments are now standard components in psychoeducational evaluations within schools, providing valuable data into why an individual may be experiencing difficulties reading at the connected text level. Results from this study draw attention to the complexity surrounding text-level readings skills. Future practice implications include the examination of both rate and accuracy by school personnel within the context of assessment and intervention efforts. However, scant attention has been paid to examining DDH and the impact of intense interventions that target phonological awareness and RAN together. Future research should examine how these deficits are impacted across time by intervention in the context of the DDH groups. Without temporal stability being established through longitudinal studies, the DDH classification scheme may have limited applied value in current assessment and intervention practice.

Conclusion

Students with dyslexia often perform poorly on phonological awareness and RAN tasks, and struggle with reading at the text level compared with their peers. The DDH theory and resulting subgroups provide a lens for conceptualizing the etiology of dyslexia (Wolf & Bowers, 1999). At the time it was proposed, the DDH theory made an important contribution to the literature by drawing attention to the importance of RAN in the reading process. The current study utilized a clinical sample of elementary age children with dyslexia, which is important given these readers are the primary target population for the DDH theory, and confirmed the results from prior studies that students with DDs perform significantly worse at the text-level (Katzir et al., 2008; Kirby et al., 2003, 2010; Lovett et al., 2000; Schatschneider et al., 2002; Spector, 2005; Wolf & Bowers, 1999; Wolf et al., 2002). However, this study also revealed instability in the reading patterns and group membership of DDH groups across time. Thus, more research is needed to understand and confirm the reliability of the DDH groups before they may be used to inform assessment and intervention efforts.

Footnotes

Acknowledgements

We express our sincere appreciation to Melissa Robinson, Ashley Breazeale, Emily Lewis, Hannah Manning, and Kelsey Walker for their assistance in collecting the data, to Drs. Kathryn Howell, Robert Cohen, and Randy Floyd, who served as thesis committee members and provided valuable feedback regarding this project, and to the students and teachers who participated in this research project.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

References

American Psychiatric Association. (2013). Diagnostic and statistical manual of mental disorders (5th ed.). American Psychiatric Publishing.

Araújo

Reis

Petersson

K. M.

Faísca

(2015). Rapid automatized naming and reading performance: A meta-analysis. Journal of Educational Psychology, 107(3), 868–883. https://doi.org/10.1037/edu0000006

Chall

(1996). Stages of reading development. Harcourt-Brace.

Cohen

(1988). Statistical power analysis for the behavioral sciences (2nd ed.) Erlbaum.

Compton

D. L.

Defries

J. C.

Olson

R. K.

(2001). Are RAN-and phonological awareness-deficits additive in children with reading disabilities? Dyslexia, 7(3), 125–149. https://doi.org/10.1002/dys.198

Ehri

L. C.

McCormick

(1998). Phases of word learning: Implications for instruction with delayed and disabled readers. Reading & Writing Quarterly: Overcoming Learning Difficulties, 14(2), 135–163. https://doi.org/10.1080/1057356980140202

Faul

Erdfelder

Lang

A.-G.

Buchner

(2007). G*Power 3: A flexible statistical power analysis program for the social, behavioral, and biomedical sciences. Behavior Research Methods, 39, 175–191. https://doi.org/10.3758/BF03193146

Katzir

Kim

Y.-S.

Wolf

Morris

Lovett

M. W.

(2008). The varieties of pathways to dysfluent reading: Comparing subtypes of children with dyslexia at letter, word, and connected text levels of reading. Journal of Learning Disabilities, 41(1), 47–66. https://doi.org/10.1177/0022219407311325

Kirby

J. R.

Georgiou

G. K.

Martinussen

Parrila

(2010). Naming speed and reading: From prediction to instruction. Reading Research Quarterly, 45(3), 341–362. https://doi.org/10.1598/RRQ.45.3.4

10.

Kirby

J. R.

Parrila

R. K.

Pfeiffer

S. L.

(2003). Naming speed and phonological awareness as predictors of reading development. Journal of Educational Psychology, 95(3), 453–464. https://doi.org/10.1037/0022-0663.95.3.453

11.

Kuhn

M. R.

Schwanenflugel

P. J.

Meisinger

E. B.

Levy

B. A.

Rasinski

T. V.

(2010). Aligning theory and assessment of reading fluency: Automaticity, prosody, and definitions of fluency. Reading Research Quarterly, 45(2), 230–251. https://doi.org/10.1598/RRQ.45.2.4

12.

Lovett

M. W.

Steinbach

K. A.

Frijters

J. C.

(2000). Remediating the core deficits of developmental reading disability: A double-deficit perspective. Journal of Learning Disabilities, 33(4), 334–358. https://doi.org/10.1177/002221940003300406

13.

Melby-Lervåg

Lyster

S.-A. H.

Hulme

(2012). Phonological skills and their role in learning to read: A meta-analytic review. Psychological Bulletin, 138(2), 322–352. https://doi.org/10.1037/a0026744

14.

National Institute for Literacy. (2008). Developing early literacy: Report of the National Early Literacy Panel: A scientific synthesis of early literacy development and implications for intervention. https://doi.org/10.1598/RRQ.45.1.2

15.

Norton

E. S.

Wolf

(2012). Rapid Automatized Naming (RAN) and reading fluency: Implications for understanding and treatment of reading disabilities. Annual Review of Psychology, 63, 427–452. https://doi.org/10.1146/annurev-psych-120710-100431

16.

Schatschneider

Carlson

C. D.

Francis

D. J.

Foorman

B. R.

Fletcher

J. M.

(2002). Relationship of rapid automatized naming and phonological awareness in early reading development: Implications for the double-deficit hypothesis. Journal of Learning Disabilities, 35(3), 245–256. https://doi.org/10.1177/002221940203500306

17.

Schwanenflugel

P. J.

Knapp

N. F.

(2016). The psychology of reading: Theory and applications. Guilford Press.

18.

Spector

J. E.

(2005). Instability of double-deficit subtypes among at-risk first grade readers. Reading Psychology, 26(3), 285–312. https://doi.org/10.1080/02702710590967834

19.

Steacy

L. M.

Kirby

J. R.

Parrila

Compton

D. L.

(2014). Classification of double deficit groups across time: An analysis of group stability from kindergarten to second grade. Scientific Studies of Reading, 18(4), 255–273. https://doi.org/10.1080/10888438.2013.873936

20.

Swanson

H. L.

Trainin

Necoechea

D. M.

Hammill

D. D.

(2003). Rapid naming, phonological awareness, and reading: A meta-analysis of the correlation evidence. Review of Educational Research, 73(4), 407–440. https://doi.org/10.3102/00346543073004407

21.

Tabachnick

B. G.

Fidell

L. S.

(2013). Using multivariate statistics (6th ed.). Pearson.

22.

Vellutino

F. R.

Fletcher

J. M.

Snowling

M. J.

Scanlon

D. M.

(2004). Specific reading disability (dyslexia): What have we learned in the past four decades? The Journal of Child Psychology and Psychiatry, 45(1), 2–40. https://doi.org/10.1046/j.0021-9630.2003.00305.x

23.

Vuolo

Uggen

Lageson

(2015). Statistical power in experimental audit studies: Cautions and calculations for matched tests with nominal outcomes. Sociological Methods & Research, 45, 260–303. https://doi.org/10.1177/0049124115570066

24.

Vukovic

R. K.

Siegel

L. S.

(2006). The double-deficit hypothesis: A comprehensive analysis of the evidence. Journal of Learning disabilities, 39(1), 25–47. https://doi.org.ezproxy.memphis.edu/10.1177/00222194060390010401

25.

Wagner

R. K.

Torgesen

J. K.

Rashotte

C. A.

Pearson

N. A.

(2013). Comprehensive Test of Phonological Processing (CTOPP-2) (2nd ed.). Pro-Ed.

26.

Wiederholt

J. L.

Bryant

B. R.

(2012). Gray Oral Reading Tests (5th ed.). Pro-Ed.

27.

Wolf

Bowers

P. G.

(1999). The double-deficit hypothesis for the developmental dyslexias. Journal of Educational Psychology, 91(3), 415–438. https://doi.org/10.1037/0022-0663.91.3.415

28.

Wolf

Bowers

P. G.

Biddle

(2000). Naming-speed processes, timing, and reading: A conceptual review. Journal of Learning Disabilities, 33(4), 387–407. https://doi.org/10.1177/002221940003300409

29.

Wolf

Denckla

(2005). RAN/RAS: Rapid automatized naming and rapid alternating stimulus tests. Pro-Ed.

30.

Wolf

O’Rourke

A. G.

Gidney

Lovett

Cirino

Morris

(2002). The second deficit: An investigation of the independence of phonological and naming-speed deficits in developmental dyslexia. Reading and Writing: An Interdisciplinary Journal, 15(1–2), 43–72. https://doi.org/10.1023/A:1013816320290